In tech, using data analysis to improve clinic efficiency and deliver better care was front and center on Day #2.
During a Presidential Plenary on Big Data, Dr. Atul Butte (UC Health) gave an overview of the University of California Health system’s work around big data, as well as his vision for the future of evidence-based medicine as being data driven. Dr. Butte succinctly explained his vision by concluding, “Evidence-based medicine for a long time has meant expert-based medicine; I think evidence-based medicine in the future is going to be data-driven medicine.” We referenced this when we heard it live on Saturday and we take a deep dive here today to really dig into this leader’s compelling presentation. We loved all kinds of things about his talk, including the fact that he used the most advanced chat available to speakers, where those watching could speak to each other, upvote or downvote each other’s questions, etc. We appreciated how much Dr. Butte participated in chat even while he answered questions – scientific meetings will never feel so immersive unless they are kept both interactive and to some degree accessible just like ENDO has been so far.
In the afternoon, the well-respected Dr. Mark Clements (Children’s Mercy Hospital, Kansas City) gave an overview of an exciting new program at Children’s Mercy Hospital, which he dubbed the “Diabetes Rapid Learning Lab.” The initiative, which has support from the Helmsley Charitable Trust, aims to “rapidly increase the deployment of successful interventions and innovative therapies in diabetes.” During Q&A, we learned that the initiative has now been formally named the “Rising T1DE Alliance.” What a great name! Overall, the initiative aims to try new and innovative interventions (e.g., new devices, drugs, mobile health, behavioral health, and care delivery strategies), monitor those interventions as they’re piloted, and rollout those interventions broadly, if successful – very exciting!
In therapy, Day #2 of ENDO 2021 was undoubtedly highlighted by a first look at data from Lilly’s once-weekly basal insulin-FC, a fusion protein combining a novel single-chain variant of insulin with a human IgG Fc domain – the same technology used in GLP-1 Trulicity’s once-weekly formulation. Phase 2 data (n=399 participants with type 2 diabetes) showed comparable efficacy, as represented by change in A1c from baseline, and safety to daily insulin degludec. We also tuned into an eye-opening session from incretin expert Dr. Dan Drucker on how researchers can approach commercializing their discoveries, drawing from his own experiences in DPP-4 inhibitors, GLP-1, and GLP-2, and sessions on treating pediatric type 2 diabetes and a meta-analysis on GLP-1 and breast cancer.
In big picture, a nationwide study using data from the National Readmission Database found that 20.2% of adults with type 1 diabetes principally admitted to the hospital for DKA were readmitted within 30 days. Intriguingly, obesity and hyperlipidemia were associated with lower rates of hospital readmission, while discharge against medical advice, anemia, hypertension, female sex, and chronic kidney disease were associated with higher rates, potentially identifying key at-risk patient populations for future intervention.
We’ve also updated our Exhibit Hall coverage from Day #1 to include companies in diabetes/obesity therapy. See here for a look at ENDO 2021 offerings from Novo Nordisk, Lilly, Xeris, Zealand, and Pfizer. What a range of experiences …
ENDO 2021 ran through this past weekend with a number of exciting sessions on Sunday. If you missed it, make sure to check out our coverage from an exciting Day #1 and see our top highlights form Day #2 below.
ENDO 2021 Day #1 Highlights – Omnipod 5 pivotal results: TIR +2.2 hours/day for adults, +3.7 hours/day for pediatrics; +7.6 hours/day TIR for small group of T2s; two posters on Novo’s STEP program for GLP-1 semaglutide in obesity
- Diabetes Technology Highlights
- 1. “Evidence-Based Medicine for a Long Time Has Meant Expert-Based Medicine”: Dr. Atul Butte Presents a Vision for Data-Driven Medicine Using Big Data; Managing Diabetes as a Turn-Based Game (e.g., Chess)
- 2. Dr. Mark Clements Outlines Children’s Mercy Hospital’s “Diabetes Rapid Learning Lab,” an Initiative to Rapidly Try, Assess, and Deploy New Interventions
- 3. Small Interview-Based Study (n=18) Finds Patients with Type 2 Experience Challenges with At-Home Diabetes Self-Management after Hospitalization
- 4. Tidepool’s Brandon Arbiter: Device Interoperability is a Given at Tidepool, “We Don’t Ask Why, But How”
- Diabetes Therapy Highlights
- 1. Lilly’s Once-Weekly Basal Insulin-FC Delivers Comparable Safety and Efficacy to Daily Insulin Degludec in Type 2 Diabetes
- 2. Case Studies in Youth with Type 2: How to Best Differentiate Types of Diabetes and Choose the Best Treatment Option
- 3. Meta-Analysis Shows GLP-1 Receptor Agonists Likely Not Linked to Breast Cancer
- Big Picture Highlights
Diabetes Technology Highlights
During a Presidential Plenary on Big Data, Dr. Atul Butte (UC Health) gave an overview of the University of California Health system’s work around big data, as well as his vision for the future of evidence-based medicine as being data driven. Dr. Butte succinctly explained his vision by concluding, “Evidence-based medicine for a long time has meant expert-based medicine; I think evidence-based medicine in the future is going to be data-driven medicine.”
As most data scientists know, data analysis starts with gathering raw data and cleaning and standardizing that data. Dr. Butte began by explaining the challenge of doing this at the massive UC Health system. In total, the UC Health system covers ten campuses (San Francisco, Davis, Berkeley, Santa Cruz, Merced, Santa Barbara, Los Angeles, Irvine, San Diego, and Riverside) and three national laboratories (Lawrence Berkeley, Lawrence Livermore, and Los Alamos). Altogether, the system includes ~200,000 employees and ~250,000 students, generating over $13 billion in operating revenue. In 2016, The UC system and UnitedHealth teamed up to create a new per-member-per-month-based accountable care organization across five academic medical centers. According to Dr. Butte, this was the catalyst for building out the UC Health Data Analytics Platform, which combines and stores data from six UC medical schools and systems (San Diego, Riverside, Irvine, Los Angeles, San Francisco, and Davis). Across the six systems, the data warehouse includes data representing 6.7 million patients, 205 million encounters, 563 million procedures, 747 million medication orders, 2 billion lab tests and more.
In one benefit of having a centralized UC Health Data Warehouse, Dr. Butte explained that the time to compile and report quality measures has been greatly reduced. When MediCal (California’s Medicaid program) comes out with new quality measures that need to be reported, the centralized database allows teams within each location to perform a subset of the work and pool it with the other locations. According to Dr. Butte, this shortened the time to report these quality measures from months to “literally a couple days or weeks.”
In another interesting project, Dr. Butte looked at using age, race, ethnicity, and area deprivation index (ADI) to predict A1c. ADI is a measure of socioeconomic status based on income, education, employment, and housing quality at the neighborhood level. The data for ADI was based in the 9-digit ZIP code for patients, which is available for all of UC’s primary care population. Not surprisingly, patients living in areas with a higher ADI had a higher mean A1c, independent of age, sex, race, and ethnicity. A1c was about 0.4% higher for those in the highest decile ADI compared to those in the lowest decile ADI.
In an interesting comparison, Dr. Butte drew a parallel between turn-based games, such as chess, and managing diabetes. The movement space in both is finite (e.g., prescribe metformin, adjust basal insulin rate, prescribe CGM, etc.) and strategy must be adjusted based on the opponents’ response. Dr. Butte noted that computers are very good at playing turn-based games and could be helpful in evaluating which “moves” in diabetes might be optimal using data. For example, a very simple decision tree (see below) can predict whether patients would do well on metformin or be escalated to another therapy with an AUC of 0.61. In the same way many chess players use computers to explore positions and plan their next moves, data and algorithms could certainly have a role to play in helping providers identify patients at higher risk and deliver better care. Ultimately, rather than fearing computers and algorithms replacing physicians, Dr. Butte focused in on using data and physicians, working together, to deliver better care.
In the afternoon, the well-respected Dr. Mark Clements (Children’s Mercy Hospital Kansas City) gave an overview of an exciting new program at Children’s Mercy Hospital, which he dubbed the “Diabetes Rapid Learning Lab.” The initiative, which has support from the Helmsley Charitable Trust, aims to “rapidly increase the deployment of successful interventions and innovative therapies in diabetes.” Later on, during Q&A, we learned that the initiative has now been formally named the “Rising T1DE Alliance.” Overall, the initiative aims to try new and innovative interventions (e.g., new devices, drugs, mobile health, behavioral health, and care delivery strategies), monitor those interventions as they’re piloted, and rollout those interventions broadly, if successful. As San Francisco residents, the approach reminds us of the “agile approach” often associated with Bay Area software companies. The Rising T1DE approach will be applied throughout Children’s Mercy Hospital’s service area, which covers eleven clinics and over 2,000 children with diabetes (~2,150 type 1s and ~250 type 2s) in Kansas, Missouri, Iowa, Illinois, Arkansas, Nebraska, and Oklahoma. This is a very cool initiative and we’re eager to hear more about learnings from the initiative at future meetings.
In two early projects done at Children’s Mercy, the team has built models to predict risk of DKA admission and A1c rises. Based on these models, providers can better identify which patients are at higher risk and provide the appropriate intervention. Ono the DKA front, a ~600-feaure recurrent neural network demonstrated strong predictive value in a test set of ~1,500 patients. For the ten patients the model determined had the highest risk of DKA, all ten were hospitalized for DKA within 180 days. Looking at the top 25 highest-risk patients, over half (56%) experienced DKA within 180 days. This sort of risk stratification should be valuable as clinics allocate their limited time and resources to those at highest risk for adverse events. Similarly, a ~305-feature random forest model for predicting 90-day rise in A1c identified 90-day A1c changes ≥0.3% with a positive predictive value of 55.5%. This came against a background rate of “~30%.”
According to Dr. Clements, most diabetes care teams are incomplete. In order to gather feedback and improve care, Dr. Clements suggested adding two new members to the traditional diabetes care team: a Quality Improvement Specialist and a Data Scientist.
In a larger scale version of a “learning lab,” Dr. Clements spent some time discussing the T1D Exchange’s Quality Improvement Collaborative. The collaborative is an EHR data sharing platform for testing and sharing of quality improvement measures. Small changes can be implemented in individual clinics, allowing successes to then be disseminated across the network and scaled. One cool learning from the collaborative came from the University of Michigan’s Dr. Joyce Lee, who identified six key habits that improve glycemic outcomes. Dr. Lee’s results showed that participating in more of the six behaviors resulted in lower A1cs: for those who participated in 0/6 habits, mean A1c was 12%, compared to a mean A1c of 8% for those who participated in 6/6 habits. Other notable learnings from the collaborative include improving screening for depression and rate of onboarding to CGM (see Keystone 2019 presentation).
A small (n=18) study of patients and caregivers found concerning results around the skills and knowledge to effectively manage their diabetes upon discharge from a hospitalization. Dr. Grace Prince (Feinberg School of Medicine) presented results from the interview-based study on the perceived and actual needs of people with type 2 diabetes upon discharge from the hospital. Two-thirds of study participants were over the age of 65 and participants included patients, caregivers, and newly diagnosed individuals. Based on interview results, Dr. Prince’s team identified three main themes that patients experience: (i) challenges in understanding self-care in the hospital; (ii) challenges in preparing items for self-care; and (iii) challenges in administering self-care at home. By identifying challenges, Dr. Page aims to establish an education model that is sustainable and effective and that can be incorporated into future efforts to enhance the remote delivery of diabetes self-care education.
In the hospital, patients reported being overwhelmed with information and struggling to understand the larger picture of diabetes self-management. This indicates that a more comprehensive and digestible overview of diabetes self-management may be helpful. Specifically, Dr. Page suggested following a “spiral learning” process in which patients are provided with a strong overview followed by “increasingly complex sequential steps” with actionable components. Dr. Page emphasized the importance of a strong overview/foundation especially for patients leaving an in-patient setting where they were likely to receive high-level and specific information that may not be applicable or realistic in an out-patient environment.
Newly diagnosed patients reported challenges utilizing diabetes self-management tools such as BGM meters, lancets, and insulin pens. To remedy this, Dr. Page suggested a “form fits function” approach in which patients are provided with an “organizing scheme” to help understand the function of their diabetes self-care tools. Specifically, Dr. Page envisions an “organizing scheme” along the lines of a medical procedure kit in which the materials and tools are organized in the order in which they’re used.
Patients and caregivers expressed discomfort transitioning from hospital-based to at-home diabetes self-management. Specifically, themes from the study interviews included (i) challenges remembering self-care steps, (ii) feeling uncomfortable dosing insulin, and (iii) feeling fearful about injecting insulin. In addition to the “spiral learning” model proposed above, to give patients and caregivers a strong understanding of the basics of diabetes self-management, Dr. Page also recommended “repetitive goal directed practice and feedback” in which patients and caregivers receive immediate feedback on dosing calculations to help overcome stress and increase confidence in self-management skills.
Brandon Arbiter (Tidepool) took part in a mid-morning session on the benefits of provider-facing digital health platforms, highlighting the importance of device interoperability to improve management for patients and providers alike. Mr. Arbiter began his presentation by providing background on Tidepool as a non-profit working in the digital diabetes data ecosystem. Particularly as a non-profit, Mr. Arbiter expressed that Tidepool thinks about interoperability as a given, asking the question of “How [do we make things interoperable]?,” rather than “Why [should we make things interoperable]?.” From Tidepool’s perspective, allowing patients and providers to choose the devices that best fit a patient’s needs is the best way to help patients “put diabetes in the background [and] get back to living.” It is with these goals in mind that Tidepool has submitted its Tidepool Loop app to the FDA for clearance as an interoperable automated glycemic controller (“iController”). If cleared, Tidepool plans to launch Tidepool Loop with compatibility with Insulet’s Omnipod patch pump and Dexcom’s G6 CGM. Tidepool also has a commercial partnership to integrate Medtronic pumps and CGMs (announced at ADA 2019); recent comments from Medtronic suggest that interoperability continues to be perceived as complex and not everything that is “possible” can be prioritized. For now, we see the 780G and the smart pen revolution as taking center stage at Medtronic. During Q&A, Mr. Arbiter, who uses DIY Loop himself, highlighted what he considers the benefits of Tidepool Loop, including novel features such as “pre-meal mode” to reduce pre-meal blousing, as well as food-based boluses to account for foods with disparate glycemic indexes. In the commercial landscape, there is also growing momentum behind device interoperability, though there have also been notable speedbumps, as well. Currently, Tandem and Insulet both have partnerships with both Dexcom and Abbott for their AID systems (Control-IQ and Omnipod 5, respectively). However, the integration work for both companies with Abbott’s FreeStyle Libre remains ongoing and could take at least a couple years – on Tandem’s 4Q20 call, CEO John Sheridan suggested integration with FreeStyle Libre is targeted “as soon as possible in 2022.”
With Tidepool’s recently flurry of activity around FDA submission for Loop, Tidepool Uploader, the company’s data aggregation software, has gone somewhat under the radar. A 2020 study on physician well-being and burnout associated with diabetes technology found that many providers can feel overwhelmed by diabetes related data and experience burn-out due to the number of platforms and products they must navigate. In the study, use of Tidepool’s “software suite,” including its data uploader technology “was found to improve clinic workflow by providing faster availability of both pump and CGM data during patient visits,” which Mr. Arbiter emphasized has the potential to improve provider well-being. Patients certainly know how important that is! Speaking directly to providers, Mr. Arbiter said “digital health tools are here to help you not hinder you.” The ability to view data from multiple diabetes devices becomes increasingly important as the amount of data grows (from growing CGM, insulin pump, and connected pen uptake) and as remote patient monitoring and telemedicine adoption increases.
Diabetes Therapy Highlights
Dr. Juan Pablo Frias (Medical Director, National Research Institute, Los Angeles, CA) presented highly anticipated phase 2 results for Lilly’s once-weekly basal insulin FC (BIF) in type 2 diabetes, demonstrating comparable safety and efficacy to daily insulin degludec. A total of 399 participants previously treated with basal insulin and oral antidiabetics were recruited to the study – 51% were women, mean BMI was 32.2 kg/m2, duration of diabetes ~15 years, baseline A1c 8.1%, and ~30% of participants were on a sulfonylurea. Participants were randomized to either insulin degludec titrated to a fasting glucose target of 100 mg/dL or one of two dosing algorithms for BIF – BIF-A1 (FG target of 140 mg/dL) or BIF-A2 (FG target of 120 mg/dL). According to Dr. Frias, higher FG targets were used with BIF “for safety,” and further phase 2 studies will utilize standardized targets.
Diving into results, both regimens of BIF were found to give similar blood sugar control to daily insulin, as represented by non-inferiority to insulin degludec on the primary endpoint of change from baseline A1c after 32 weeks of treatment. On average, individuals on degludec achieved a 0.7% drop in A1c on degludec and ~0.6% drop on either BIF regimen. In addition, similar proportions of patients in the three arms achieved A1cs <7.0% without nocturnal hypoglycemia <54 mg/dL or ≤70 mg/dL, though higher proportions were seen for both BIF regimens.
On safety, both BIF regimens had significantly lower rates of documented hypoglycemia ≤70 mg/DL – this difference was primarily driven by nocturnal hypoglycemia benefit, although BIF trended toward a non-nocturnal hypoglycemia benefit as well. That being said, the only two cases of severe hypoglycemia were in the BIF-A2 arm, and this point was brought up by audience members during Q&A. Dr. Frias clarified that both patients were treated by a third party by oral carbohydrates, and no long-term sequelae were apparent. Dr. Frias also noted that in general, incidence of hypoglycemia was higher across all three arms due to patients wearing unblinded CGM throughout the study, so ascertainment of hypoglycemic events was particularly robust. Excitingly, CGM results will be shared at the upcoming ADA meeting. To build upon these promising phase 2 results, BIF is currently progressing through “a large phase 2 program exploring the efficacy and safety of BIF in a broad patient population including type 1 diabetes” with what we believe to be a unified FG target of 100 mg/DL. Through a search on ClinicalTrials.gov, we identified a phase 2 (n=264) in insulin-naïve patients with type 2 diabetes and a phase 2 (n=254) in individuals with type 1 diabetes, both expected to complete in September 2021.
Dr. Frias’s also shared BIF’s impressive phase 1 pharmacokinetic data, which Lilly management referred to as “remarkable” in 2019, without sharing any data at that time. Indeed, BIF seems to have a very attractive pharmacokinetic profile, achieving a flatter peak-to-trough profile over six weeks than daily insulin degludec.
As explained by Dr. Frias, once-weekly basal insulins have been of high interest due to their potential to improve compliance and adherence to insulin therapy and prevent therapeutic inertia. BIF is a fusion protein that combines a novel single-chain variant of insulin with a human immunoglobulin G (IgG) Fc domain – this is the same technology used for Trulicity’s once-weekly formulation. Of note, Dr. Frias again emphasized that BIF can be co-formulated with weekly incretins; although a “next gen” fixed-ratio combination GLP-1/basal insulin with dulaglutide/BIF comes top-of-mind, we wonder if co-formulation with Lilly’s phase 3 tirzepatide may also be on the table. Basal/GLP-1 combinations annualized at just $558 million in 2020, though both Sanofi’s Soliqua and Novo Nordisk’s Xultophy delivered double digit YOY Growth (+26% and +13% at CER, respectively).
Elsewhere in the once-weekly basal insulin landscape, Novo Nordisk released full positive phase 2 results for insulin icodec at ADA 2020. Icodec’s expansive phase 3 ONWARDS program was launched in 4Q20 and will extend through 2022. As some KOLs like Dr. Julio Rosenstock (Dallas Diabetes Research Center) have expressed that icodec is likely to be more transformative in patients with type 2 diabetes, given less flux in insulin need, we’re eager to see how both companies view the drug’s potential in type 1 diabetes.
Through the lens of multiple case studies, Drs. Philip Zeitler (University of Colorado School of Medicine) and Petter Bjornstad (Children’s Hospital Colorado) outlined best practices for treating type 2 diabetes in youth. Recommendations focused on initial approaches to evaluate and manage new-onset diabetes, select pharmacotherapies, and manage complications and cardiovascular risk. Prior to discussing cases, Dr. Zeitler emphasized the importance of effectively treating type 2 in youth, as there is an inverse relationship between age of type 2 onset and complication risk/mortality.
When treating someone with new-onset diabetes, it is important to determine if the patient has type 1 or type 2 based on pancreatic autoantibodies. No clinical characteristic fully excludes type 1 diabetes, type 1 diabetes is seen more often than type 2 in youth, and obesity is common in both type 1 and 2 diabetes. However, when a patient presents with A1c over the ADA and ISPAD target of 7%, it is important to start the patient on once-daily basal insulin because initial metformin monotherapy is associated with a four to 10-fold risk for loss of glycemic control due to the course of treatment not being as aggressive as needed. We also note that Novo Nordisk’s GLP-1 Victoza was approved for individuals aged 10 to 17 in June 2019.
If someone presents with severely elevated fasting triglycerides beginning fenofibrate treatment is the best next step. Dr. Bjornstad also discussed the relationship between bariatric surgery and DKD based on data from the TODAY and Teen-LABS trials. He stated that the mechanisms by which surgery attenuates DKD are unclear, which is further complicated by the unknown nature of how much more effective bariatric surgery is to new therapies like SGLT-2 inhibitors and GLP-1 agonists. If the patient already on metformin and an ACE inhibitor presents with albuminuria instead, then Dr. Bjornstad recommends adding an SGLT-2 inhibitor to prevent progression to macroalbuminuria and end-stage kidney disease. However, this is only a feasible treatment option if the youth is 18 years or older, as SGLT-2 inhibitors are not yet approved for pediatric use.
3. Meta-Analysis Shows GLP-1 Receptor Agonists Likely Not Linked to Breast Cancer
Dr. Giovana Fagundes (Hospital de Clinicas Porto Alegre) presented the findings of a systemic review and meta-analysis to determine if using GLP-1s is related to breast cancer, based on the findings of an imbalance of breast neoplasms between groups in the SCALE trial of liraglutide. 52 trials total on liraglutide, semaglutide, albiglutide, exenatide, dulaglutide, and liraglutide were used to assess if there is any statistical association between GLP-1 agonist use and breast neoplasms (baseline A1c: 5.5-9.1%; BMI: 25.3-39.3 mg/m2). Subgroup analyses found no difference in breast cancer incidence or benign breast neoplasms between groups. However, the meta-analysis was limited by the low number of breast neoplasms in the data set, lack of standardization in breast neoplasm reports between the trials, and lack of hormone data (e.g., menopausal status, hormone replacement therapy usage, or age of breast neoplasm occurrence). During Q&A, an attendee brought up the potential relationship between breast cancer and obesity. Dr. Fagundes shared that while her review did not show a statistically significant difference between weight and breast cancer incidence, maybe longer studies can evaluate whether participants who lost different amounts of weight while using a GLP-1 agonist showed differences in breast neoplasm incidence.
Big Picture Highlights
1. One-Fifth of DKA Hospitalizations in Type 1s Result in Hospital Readmission within 30 Days; Readmission Associated with Two Times Higher Mortality
During the Sunday afternoon session on “Improving Diabetes Care,” Dr. Hafeez Shaka (John H. Stroger Jr. Hospital of Cook County) presented results from a nationwide study of diabetic ketoacidosis readmissions in adults with type 1 diabetes. Dr. Shaka’s study used the National Readmission Database for all adult type 1s who were principally admitted for DKA between January 1 and November 30, 2017. In total, the database recorded 91,625 DKA hospitalizations. Of those discharged alive (n=91,401), the rate of readmission to the hospital within 30 days was 20.2% (18,553 readmissions) – meaning a staggering one in five patients were readmitted. A majority of the reasons for readmission were also for DKA.
Hospital readmission within 30 days was associated with a two times higher mortality, compared to the original DKA hospitalization (HR=2.06, p<0.001). Additionally, the readmission was associated with a longer mean length of stay (by one-day) and total hospitalization cost (by $8,217).
Interestingly, obesity and hyperlipidemia were associated with lower rates of hospital readmission. Obesity was associated with a 30% reduced risk for readmission (HR=0.7, p<0.001) and hyperlipidemia was associated with an 8% reduced risk (HR=0.92, p=0.007). On the other side, discharge against medical advice (HR=1.54), anemia (HR=1.42), hypertension (HR=1.28), female sex (HR=1.14), and chronic kidney disease (HR=1.13) were associated with higher rates of hospital readmission.
For a health economics perspective, the estimated cost of DKA-related hospitalizations in the US was about $5 billion in 2014. In 2014, a total of 188,965 DKA-related hospitalizations were recorded with an average cost of $26,556 per hospitalization (Diabetes Care, 2018). (Note that the Diabetes Care figures include people with both type 1 and type 2 diabetes.) On the promising side, recent nationwide data from the UK suggests use of CGM – FreeStyle Libre, in this case – can reduce DKA events and hospitalizations by 80% (see presentations at ADA 2019 and a paper in Diabetolgia in 2019).
Following Sunday’s presentation, our associates participated in Monday morning’s Press Conference to further discuss the results with Dr. Shaka.
Q: Was it surprising to see obesity and hyperlipidemia being associated with lower rates of readmission for DKA?
Dr. Shaka: That’s a very good question. I do a lot of obesity-based research, and you may be familiar with certain concepts that have come up recently – one would be the obesity paradox that has been reported in hospitalized patients for various conditions, including strokes, HF, etc. The other concept is what is called “metabolically healthy obesity.”
However, in this population of type 1 diabetes, I think the reason why obesity seems to be protective in this case is because obesity might be a surrogate marker for patients who have been traditionally more compliant with insulin. Patients with type 1 diabetes need insulin to store body fat. Most of these patients who are poorly controlled tend to be either underweight or within the normal limit. Patients who have been more compliant with insulin are the ones who tend to be more overweight and have other lipid disorders, so that is a possible hypothesis, but it remains to be really tested to see if there’s another pathophysiologic basis.
Q: Why might females be more likely to be readmitted?
Dr. Shaka: That’s a good question. There’s some belief that females tend to be more health-seeking for disease severity when compared to the average male. I have not seen a specific study on type 1 diabetes to see if this also applies to this cohort, but that could be one of the reasons. It could be that females tend to present more when they feel like they are going into DKA compared to males. Even after adjusting for age, disease comorbidities, and hospital factors – sex remained the strongest positive predictor. This is also consistent with other studies that have been done on DKA. That brings up an interesting research topic, which would be to assess if there is a difference in outcomes between males and females during hospitalizations during DKA. That’s something my research team will be working on
Q: Is the 20% readmission rate surprising? Is it similar to previously reported rates?
Dr. Shaka: The 20% readmission rate was quite surprising, and it’s quite high by any metrics – that’s one in every five patients just looking at a 30-day window. That means if you extend beyond that, you are averaging almost one or two readmissions per patient per year, which is something that is not talked about enough since we’re usually all about glucose control. Now, this patient population might be acuter because you have other medical conditions. There was another study done on all DKA, which reported a readmission rate of ~12 to 13%. When you compare it to studies for other chronic conditions such as HF, this amount is quite alarming.
Q: What was the average total hospitalization cost for admission and readmission? Do you have data on patients with type 1 vs. type 2 diabetes?
Dr. Shaka: In our current study, we did not compare type 2 diabetes patients on the cost of hospitalization vs. type 1 diabetes. This was impossible because the age group difference in patients who are traditionally type 1 vs. type 2 diabetics was so large that in-between, you have way higher comorbidities, which significantly contributes to their length of stay and cost of hospitalization. So, a direct head-to-head comparison between type 1 and type 2 diabetes was not done in this study. We only compared how much it cost hospital during the initial hospitalization vs. readmission. We found out the readmission was much higher in cost. The goal is to actually prevent all costs associated with readmissions, or a significant portion of it. One of the things that was not highlighted here was the total cost of readmissions in the US population within that year, which ran into millions of dollars. So, if we can cut down just a fraction of that, we can not only save patients from being hospitalized but the 1000s of days of lost productivity, as well as decrease the cost on the overall healthcare system.
--by Hanna Gutow, Katie Mahoney, Ursula Biba, Rhea Teng, Albert Cai, and Kelly Close